Understanding AI Topic Tagger

Prev Next

Silhouette of a person standing, representing human figures in various contexts. Business admin

Diamond.png Any Gong plan*

Admin Center > Agent Studio

AI Topic Tagger is an AI agent that helps your team identify and analyze business-critical topics discussed in customer conversations. It surfaces recurring subjects across calls, such as pricing, next steps, or product issues, so you can coach teams, track messaging, and focus on what matters most.

The list of topics associated with your calls is unique to your company. We go through your calls, flag certain words and phrases, and associate them with specific topics. Topic models can be created for each workspace in your organization, and in any language that has transcript support.

Topics appear across Gong: on the call page, in filters, in team analytics, and in Whisper, giving you insight into patterns and behaviors across your sales org.

Understanding topics and topic models

A topic is a subject that Gong AI identifies as being frequently discussed in your company’s calls. Topics help you understand what conversations are happening on the ground, how they evolve over time, and who is driving them.

For example:

  • “Pricing” may surface when prospects say things like “We need a discount.”

  • “Call setup” may capture phrases like “Let’s start the recording.”

  • “Next steps” might be triggered by “Let’s connect next week.”

Each topic is powered by a topic model, which determines how topics are generated.

Base topic model

This is an out-of-the-box model  that automatically detects commonly used phrases. Topics in the base model include:

  • Call Setup: For example “zoom links” or “start the recording”

  • Small Talk: For example “it's raining” or “where are you based?”

  • Next Steps: For example “next week” or “let’s set up another call”

  • Pricing: For example “that’s too expensive” or ”we need a discount”

  • Wrap-up: For example “thank you very much guys” or “ have a great day”

The base model topic is displayed in English, and enabled by default for everyone.

Custom topic models

Custom topic models generate topics that reflect your organization’s unique language, teams, or product lines. These models allow you to surface insights that matter most to your business.

You need about 1000 hours of recorded calls in a workspace before you can create a custom topic model. The model is created in collaboration with your Gong CSM. Once it’s ready, you can activate and deactivate the model on your own, and edit the topic names.

Create a custom topic model

Before creating a custom topic model you’ll need:

  • At least 1,000 hours of recorded calls in a workspace

  • Collaboration with your Gong CSM to define segmentation and filters

To create a custom topic model:

  1. Once the topic model is defined, your CSM will run the topic model. You’ll receive an email when it’s ready. This takes about 10 minutes.

  2. Once the model is ready, click Admin Center > Agent Studio.

  3. Hover over AI Topic Finder and click Settings.

  4. Click the name of the new topic model. It won’t be activated yet. The model will open, with all of its topics listed along the left side of the screen.

  5. Click a topic name to open a panel on the right that shows what this topic is based on.

  6. At the top of the panel, you’ll see an AI-suggested name and description. Click in the fields to edit.

  7. In the Display section, choose whether to display the topic and what color to use for it. Changes are automatically saved.

  8. When you’re satisfied with the topic names and descriptions, go back to the Topic model page and click Activate. It may take a few hours for the model to be activated.

Common phrases: This section shows which phrases came up in this topic.

Call snippets: This section contains the snippets that were used for this topic. You can listen to the snippets and show/hide the transcript. 

Unnamed topics: By default, unnamed topics aren’t displayed in Gong. You can listen to them, enter a name, and toggle on to display them. 

Topics that have similar content: If there are topics with similar content, you can merge them by assigning the same name to them.

Edit topic names

You can edit topic names any time, and there’s no need to reactivate the model afterwards.

  1. Click Admin Center > Agent Studio.

  2. Hover over AI Topic Tagger and click Settings.

  3. Click the name of the topic model you want to edit. The model will open, with all of its topics listed along the left side of the screen.

  4. Click a topic name to open a panel on the right that shows what this topic is based on.

  5. At the top of the panel, you’ll see an AI-suggested name and description. Click in the fields to edit. You can also change the color associated with the topic.

  6. Changes are automatically saved and displayed when this topic comes up in your calls.

Where can your team see topics?

Topics are shown in the following places in Gong:

On the call page

Topics on the call page show which subjects were covered during a call. Team members can click directly into the recording where the topic was covered. The subject discussed the most is referred to as the main topic.

As a call filter

Team members can filter calls according to call topics, so they focus on calls referring (or not referring) to topics that matter to them. For example, they can filter for calls in which the topic pricing was found, or which the topic Objections was not found.

In team stats

Managers can help their teams learn and grow by working with topic statistics. For example:

  • Review the time spent by each team member on promoting a new value proposition

  • Verify that team members are not spending too much time on small-talk

  • See what topics are being discussed more by A-level reps

In Whisper

Topics can be used when setting up coaching Whispers.

Build a custom topic model strategy

Here are some guidelines for considering what types of topic models you want to build.

For larger enterprises consider:

  • Using a department-based approach to create separate models for SDRs, upmarket AEs, marketing strategists, and sales enablement teams

  • Using a business-based approach to create separate models that cover lost deals, new upmarket opportunities, mid-market sales wins, and conversations you want to avoid or repeat

  • Segmenting models for teams that support different product lines

For smaller and medium businesses consider:

  • Creating models for pre-sales and post-sales teams

*The features available to you depend on your company’s plan and your assigned seats.